Page 2 of 2

Re: Downloadable trained neural net for Go

Posted: Wed Apr 13, 2016 4:05 am
by Krama
I noticed it can't read out ladders so you might want to play a ladder breaker and win :)

Re: Downloadable trained neural net for Go

Posted: Fri May 20, 2016 7:56 am
by wauske
wauske wrote:Well, you guys are a lot stronger than I am so I have yet to win :clap:

Luckily I am a lot more tech-savvy than I am a Go player. I noticed that the site only offered the net.js as a download, not the other pre-requisite files. So, I downloaded them in their original format and zipped it. For those wanting to play with it offline or on their own (private) webspace, here is the download link: http://millsf.home.xs4all.nl/Deepgo.zip

This is a direct, 1-on-1 download of the page. All credit goes to its creator.

To use it: unzip it to any directory and open the index.html (Chrome or Edge work fine, IE11 doesn't). I also uploaded it to a private webserver which works too.

I'll also ask the owner for permission, should he decide not to grant it I'll remove the link right away.
Download has been updated.
Added functionality is that when opening the website on an Android Phone with Chrome 39 or newer you can choose "Add to home screen" from the menu. As far as I know this only works when running from webspace and not from internal storage but i haven't tried running it from internal storage.
The website will then behave like an app and appears to stay in memory longer which means you can open other apps and return to the DCNN without losing your game.

This is mostly experimental and I won't vouch for it working on every Android Phone but it works on my Moto X but doesn't from my Galaxy Tab 2 my guess would be that memory is the cuplprit.

Re: Downloadable trained neural net for Go

Posted: Fri May 20, 2016 10:08 am
by Krama
Would someone be willing to take any free go software that is open sourced and embed this neural network to it's MCTS?

I am wondering how strong can such a basic thing become.

Re: Downloadable trained neural net for Go

Posted: Fri May 20, 2016 11:26 am
by yoyoma
Krama wrote:Would someone be willing to take any free go software that is open sourced and embed this neural network to it's MCTS?

I am wondering how strong can such a basic thing become.
Someone already added some limited DCNN capability to Pachi -- only the root node. Pasky is not actively developing Pachi anymore but he is supporting pull requests.

https://github.com/pasky/pachi/pull/28